Latgalistikys kongresu materiali, III. 2011. - Latvijas Universitāte

for administration purposes, while schooling is through Russian; Russian
is therefore by far the preferred code when speaking with children.
Latgalian is still used with elderly people, and we may infer that it was
once necessary to speak with the people of the neighbouring villages as a
mesolect. This could explain the relatively strong position enjoyed by
Latvian in the past, since the two codes could easily overlap (for a Russian
speaker). Note however that (apart from Latgalian) the shifts are small in
absolute terms, despite the strong orange colour code assigned to the
geographical unit; while once again the same type of arrow has been
assigned to both Latvian and Latgalian, despite the difference in starting
conditions and in the extent of change for each of the codes. In other
words, to facilitate accurate judgement and analysis, we require a method
to measure the extent of the change, and not just the direction.
4. A further step in analysing the linguistic evolution of a territory
involves calculating the difference between the highest and lowest perceived
usage values of a given code in relation to the generational variables
(elderly relatives, partner, children). This allows us to quantify
variations in the linguistic usage of the different generations or in the interchanges
between the different generations (as in this study): low values
indicate similar usage of the various codes across the different generations,
while higher values indicate that linguistic usage becomes increasingly
more diverse as a function of the age of the speaker, although this relationship
is not necessarily linear (the trend may be temporarily inverted with a
higher or lower value in the middle generation not shown by this type of
analysis). See maps 4, 5, 6 for an example.
Light shadings represent basically stable situations, that is, situations
where there is little difference between the highest and lowest value; dense
shadings depict situations where there is significant change in linguistic
use across the generations. Of course these charts do not tell us which
codes are actually being used, on what occasions and by which generations:
what they provide is a general picture of the difference in declared
linguistic usage across the demographic groups or the situations under
analysis.
However, since the values thus obtained are not weighted in relation
to the level of usage of the various codes in the different municipalities,
the output charts may not provide a completely accurate picture of intergenerational
variation. In other words, absolute difference values do not
show us whether the language is more or less spoken in real situations. To
understand this concept by analogy, let us imagine a political party that
101